카드거래정보를 활용한 맛집 추천 기법

다국어 초록 (Multilingual Abstract)

With the spread of the Internet and the development of the IT industry, much information can be obtained, while users have difficulty in selecting useful information. As a result, the selective information distribution method has emerged and personalized recommendation services have emerged. Collaborative filtering is the most successful method of recommendation systems, and association rules are a common and traditional method in the recommendation. Personalized recommendations have made it possible for users to customize their services. However, the restaurant recommendation system is only recommended through the selective information distribution method. As a result of the survey, users showed the most trust in recommendation of nearby people when they visited restaurants. Therefore, it is necessary to provide a personalized restaurant recommendation service through the information of actual people's visit history.
In this paper, after analyzing basic data using restaurant transaction data for 3 months, we apply various similarity in collaborative filtering algorithm and compare the accuracy after recommendation. We also compare recommended restaurants by association rules and blogs. This study will help us to develop various strategies to respond to customer needs in various fields base on the real transaction data.

목차 (Table of Contents)

• I. 서 론 1
• 1.1. 연구 배경 및 필요성 1
• 1.2. 연구 목적 및 범위 4
• II. 관련 연구 5
• 2.1. 협업필터링 5
• 2.1.1. 사용자 기반 협업필터링 6
• 2.1.2. 아이템 기반 협업필터링 6
• 2.1.3. 유사도 7
• 2.1.4. 선호도 예측값 9
• 2.2. 연관규칙 9
• 2.2.1. 연관규칙분석의 척도 10
• 2.2.2. 연관규칙분석의 절차 11
• III. 데이터 12
• 3.1. 데이터 설명 12
• 3.2. 데이터 특징 13
• 3.3. 데이터 정제 17
• 3.3.1. 차원 축소 17
• 3.3.2. 데이터 분할 18
• IV. 협업필터링을 이용한 추천 19
• 4.1. 사용자 기반 협업필터링 기법 20
• 4.1.1. 데이터 변환 20
• 4.1.2. 이웃 개수 선정 20
• 4.2. 아이템 기반 협업필터링 기법 22
• 4.3. 협업필터링 결과 비교 23
• Ⅴ. 연관규칙을 이용한 추천 25
• 5.1. 연관규칙 선별 25
• 5.2. 선호 점수 계산 27
• 5.3. 연관규칙 추천 결과 28
• Ⅵ. 결론 31
• 참 고 문 헌 33